OpenAI's top models now run on Amazon's cloud. GPT-5.5, GPT-5.4 and the Codex AI agent are on AWS Bedrock as of 1 June 2026 — and the price is identical to OpenAI's own.
What landed on Bedrock
Three OpenAI products are now generally available through Amazon Bedrock: GPT-5.5, OpenAI's most advanced frontier model; GPT-5.4; and Codex, the coding agent. Each runs on Bedrock's next-generation inference engine and inherits the enterprise controls AWS customers already use — IAM access management, encryption, VPC isolation, and CloudTrail logging. For a regulated buyer, inheriting those controls matters as much as the model.
The commercial terms are the news. According to Amazon, pricing matches OpenAI's direct rates with no AWS surcharge, and the usage draws down a customer's existing AWS spending commitment. A company that has committed millions to AWS can now spend part of the commitment on OpenAI models — consolidating AI and cloud onto one bill instead of two.
💡OpenAI on AWS, in brief
GPT-5.5, GPT-5.4, and Codex generally available on Amazon Bedrock (1 June 2026) · pricing matches OpenAI's direct rates, no AWS fees · usage counts against existing AWS commitments · runs on Bedrock's next-gen inference engine with IAM, encryption, VPC isolation, and CloudTrail logging.
The Azure question
OpenAI and Microsoft Azure have been near-synonymous since 2023, so OpenAI's frontier models showing up on Amazon's Bedrock is a genuine widening of distribution. Whatever the contractual fine print, the practical reality on 1 June is plain: a developer who lives on AWS no longer has to leave AWS to use GPT-5.5. The wall between the two biggest clouds, where OpenAI is concerned, is lower than it was a week ago.
The strategic logic runs both ways. For OpenAI, every cloud that carries the models is another distribution channel and another hedge against depending on one partner. For AWS, carrying GPT-5.5 closes a gap — Amazon's own models and its Anthropic stake left a hole exactly where the most-requested models sat. Putting OpenAI on Bedrock fills the hole and keeps AWS customers from drifting to Azure for the one thing AWS could not previously sell.
The model is no longer the moat. When the same GPT-5.5 lands on every major cloud at the same price, the competition moves to distribution, tooling, and trust — and the contest becomes about who is easiest to build on, not who is smartest.
Distribution is the real contest
A pattern is hardening across the year's AI-industry news: the frontier model is becoming a component, and the contest is shifting to who can put the component in front of the most builders with the least friction. Selling GPT-5.5 through Bedrock at parity pricing is a distribution move, not a technology one — and distribution, historically, is what decides platform fights long after the technology stops being the differentiator.
There is a thread here that connects to the bigger argument. As frontier models commoditise into interchangeable components, the values embedded in each model — what it will and will not do, whose dignity it protects — risk being flattened into a pricing footnote. Emergent Intelligence, the frame I use for these systems, insists the choice of which intelligence you build on is never only a procurement decision. On a parity-priced Bedrock menu, that choice gets easier to make carelessly.
For most buyers, though, 1 June is simply good news: more choice, the same price, one less reason to run two clouds. The model wars made the headlines. The distribution wars will decide the outcome.
Frequently Asked Questions
These are the questions developers, cloud architects, and AI-industry readers have been asking since the Bedrock availability went live. Short answers follow, drawn from OpenAI's and Amazon's announcements.
What is the OpenAI–AWS Bedrock availability?
In short, OpenAI's GPT-5.5, GPT-5.4, and Codex are now generally available through Amazon Bedrock as of 1 June 2026. The answer, simply put, is that AWS customers can call OpenAI's frontier models without leaving Amazon's cloud. The key is the pricing: data from Amazon shows the rates match OpenAI's direct pricing with no AWS surcharge.
How does running OpenAI models on Bedrock work?
According to Amazon, the models run on Bedrock's next-generation inference engine and inherit AWS enterprise controls — IAM access management, encryption, VPC isolation, and CloudTrail logging. The evidence shows usage counts against a customer's existing AWS commitments, which means AI spending consolidates onto the same bill as the rest of a company's cloud workloads.
Why is this significant for the cloud market?
The answer is distribution. OpenAI has been tied to Microsoft Azure since 2023, so its frontier models appearing on AWS Bedrock widens where the models can be used. Analysis shows the move turns the frontier model into a component available across clouds at parity pricing — which shifts the competition from who has the smartest model to who is easiest to build on.
Who is this for?
This is for the enterprises and developers already building on AWS who wanted OpenAI's models without running a second cloud. In other words, the change targets regulated and large-scale buyers who value AWS's security controls and consolidated billing. According to Amazon, early adopters include teams building production-scale AI applications on the combined stack.
What are the risks, and what does it mean for Azure?
Analysis demonstrates a few things to watch. First, the contractual status of OpenAI's Azure relationship is not spelled out in the AWS announcement, so the exact exclusivity picture stays partly unclear. Second, parity pricing makes switching models easy, which can flatten the real differences between them into a footnote. Third, for Microsoft Azure, OpenAI on a rival cloud is a competitive loss of a once-exclusive draw — evidence that even the deepest AI partnership is now multi-cloud.
